Automatic visual and acoustic analytics for event detection

Patent metadata

Field	Value
Publication number	US-2017357233-A1
Application number	US-201715596557-A
Country	US
Kind code	A1
Filing date	May 16, 2017
Priority date	Jun 8, 2016
Publication date	Dec 14, 2017
Grant date	—

Abstract

Systems and methods are provided for detecting events in industrial processes. An acquisition system may include one of a camera and an audio recorder to acquire monitoring data in the form of one of imaging data and acoustic data, respectively. A computer system, may include a machine learning engine and may be programmed to classify the monitoring data under a classifier, quantify, based on the classifier, the monitoring data with at least one quantifier, and detect an event when the at least one quantifier satisfies a predetermined rule corresponding to the at least one quantifier.

First claim

1 . A method, comprising:
acquiring monitoring data from an acquisition system; classifying the monitoring data under a classifier; quantifying, based on the classifier, the monitoring data with at least one quantifier; and detecting an event when the at least one quantifier satisfies a predetermined rule corresponding to the at least one quantifier.

2 . The method of claim 1 , further comprising associating the event with a probability of being true.

3 . The method of claim 1 , wherein the monitoring data is one of imaging data and acoustic data.

4 . The method of claim 1 , further comprising analyzing historical data to generate a library of expected events, wherein each of the expected events is associated with one of an alert and a control signal.

5 . The method of claim 4 , wherein at least one of the classifying, the quantifying, the detecting, and the analyzing is performed by a machine learning engine operated by a computer system.

6 . The method of claim 4 , further comprising, when an event is detected:
matching the detected event to one of the expected events; if there is a match, transmitting one of the alert and the control signal; and if there is no match, generating and transmitting another alert.

7 . The method of claim 6 , wherein one of the alert and the another alert is transmitted to an operator of an equipment related to the detected event.

8 . The method of claim 7 , further comprising adapting protocols of the machine learning engine based on a feedback from the operator.

9 . The method of claim 6 , wherein the control signal is transmitted to a controllable equipment related to the detected event.

10 . The method of claim 6 , further comprising, augmenting the historical data by:
historizing the acquired monitoring data; and when an event is detected,
historizing the detected event,
associating the detected event with the acquired monitoring data;
if there is a match,
historizing one of the alert and the control signal, and
associating one of the alert and the control signal with the detected event, and

if there is no match,
historizing the another alert, and
associating the another alert with the detected event.

11 . A system, comprising:
an acquisition system to acquire monitoring data; and a computer system including a machine learning engine programmed to:
classify the monitoring data under a classifier;
quantify, based on the classifier, the monitoring data with at least one quantifier; and
detect an event when the at least one quantifier satisfies a predetermined rule corresponding to the at least one quantifier.

12 . The system of claim 11 , wherein the acquisition system includes a camera and the monitoring data is imaging data.

13 . The system of claim 11 , wherein the acquisition system includes an audio recorder and the monitoring data is acoustic data.

14 . The system of claim 11 , wherein machine learning engine is further programmed to analyze historical data stored in a database to generate a library of expected events, wherein each of the expected events is associated with one of an alert and a control signal.

15 . The system of claim 14 , wherein, when an event is detected, the computer system is programmed to:
match the detected event to one of the expected events; if there is a match, transmit one of the alert and the control signal; and if there is no match, generate and transmit another alert.

16 . The system of claim 15 , wherein the computer system is programmed to transmit one of the alert and the another alert to an operator of an equipment.

17 . The system of claim 15 , wherein the computer system is programmed to transmit the control signal to a controllable equipment.

18 . The system of claim 15 , wherein the computer system is further programmed to augment the historical data in the database by:
historizing the acquired monitoring data; and when an event is detected,
historizing the detected event,
associating the detected event with the acquired monitoring data;
if there is a match,
historizing one of the alert and the control signal, and
associating one of the alert and the control signal with the detected event, and

if there is no match,
historizing the another alert, and
associating the another alert with the detected event.

19 . The system of claim 14 , wherein the computer system is further programmed to:
query the historical data from the database; and display the queried data on a display unit to allow for identification of unexpected events.

20 . A non-transitory machine-readable medium storing instructions adapted to be executed by one or more processors to perform operations comprising:
acquiring monitoring data from an acquisition system; classifying the monitoring data under a classifier; quantifying, based on the classifier, the monitoring data with at least one quantifier; detecting an event when the at least one quantifier satisfies a predetermined rule corresponding to the at least one quantifier; analyzing historical data to generate a library of expected events, wherein each of the expected events is associated with one of an alert and a control signal; matching the detected event to one of the expected events; if there is a match, transmitting one of the alert and the control signal; and if there is no match, generating and transmitting another alert.

Assignees

• Exxonmobil Res & Eng Co

Inventors

• Gurciullo Christopher S

Classifications

• G06N20/00 Primary

  Machine learning · CPC title

• G06F16/2455

  Query execution · CPC title

• G06F16/61

  Indexing; Data structures therefor; Storage structures · CPC title

• G05B23/0218

  characterised by the fault detection method dealing with either existing or incipient faults · CPC title

• G05B2219/36039

  Learning task dynamics, process · CPC title